Denoising U-Net

Denoising U-Net

Removes noise from old gramophone recordings with a two-stage U-Net and returns a cleaned WAV file. It runs on a CPU and suits archivists, audio restorers and music historians.

Made by teamup-tech on Hugging Face. These notes are Mindket's, written from its Space page and README; the app's own files and license are on Hugging Face.

About

What it is: Denoising U-Net cleans noise out of historical gramophone recordings. You upload a recording and get a denoised WAV file back. It is built on the two-stage U-Net of Moliner and Välimäki, offered as a HARP endpoint and run on a CPU.

Who it suits: Archivists, audio restorers, music historians and hobbyists who work with old shellac and gramophone recordings. It also suits researchers who want to try this model without installing it.

Title: Who restores old recordings
Industry | Who uses it | How they use it | Example
Archives | Sound archivists | Clean digitized gramophone discs for listening copies | Denoise a transferred record before cataloguing
Music | Music historians | Make old recordings easier to hear | Listen to a rare disc with less surface noise
Audio restoration | Restorers and hobbyists | Try a neural denoiser on noisy transfers | Compare the cleaned file with the original
Research | Audio researchers | Test the two-stage model on their own material | Run the checkpoint on a sample recording

If your recordings are modern and clean, or you need stereo output, you can skip it, since the input is converted to mono.

Where to use it:

• A web browser: the Space runs on Hugging Face with nothing to install
• Your own computer, running Python 3.11 on Windows, macOS or Linux
• A HARP setup, since the Space is a HARP endpoint

Use case | Who | What they get back
Denoise a gramophone recording | Archivists | A cleaned WAV file
Try the model without installing it | Researchers | The original checkpoint running on CPU in a browser
Prepare a recording for listening | Historians | A mono file at 44.1 kHz

How to set it up:

• Python 3.11 and pip
• The Python packages in requirements.txt
• A recording to clean

1. Install the Hugging Face hub package and download the Space into a folder.
2. Move into that folder.
3. Install Gradio 6.24.0, then the packages from requirements.txt.
4. Start the app with Python.

pip install -U huggingface_hub
hf download teamup-tech/Denoising-UNet --repo-type space --local-dir Denoising-UNet
cd Denoising-UNet
pip install gradio==6.24.0
pip install -r requirements.txt
python app.py

How to use it: Open the Space in your browser, or start the local copy with the command below and open the page Gradio serves. Upload a recording and wait for the denoised WAV file. The page first folds your audio down to one channel at a 44.1 kHz sample rate.

python app.py

Example: You upload a digitized gramophone record with heavy surface noise. The app converts it to mono at 44.1 kHz, runs the U-Net on the CPU and hands you back a cleaner WAV file.

For newcomers and experts: If you are new to this, open the Space and upload a short clip from an old recording. Play the result next to the original and listen for the hiss.

If you know the field, this is the two-stage denoising U-Net of Moliner and Välimäki, built for historical gramophone material. The Space ships the original checkpoint and runs inference on CPU. The trade is a fixed pipeline: everything is made mono and resampled, so stereo detail is lost.

You upload a recording
It becomes mono at 44.1 kHz
The two-stage U-Net removes noise
You download a cleaned WAV

Good to know:

• Hardware: the Space runs on cpu-basic at Hugging Face, so it needs no GPU
• Input handling: your audio is folded to a single channel at 44.1 kHz before processing
• Weights: the original checkpoint is included in the Space
• License: mit
• Cost to run: none named
• Data it reaches: the recording you upload, which is processed in the Space

On Hugging Face, 13 November 2026

• Likes: 1
• Built with: gradio
• License: mit

Links

• Space: https://huggingface.co/spaces/teamup-tech/Denoising-UNet
• Files: https://huggingface.co/spaces/teamup-tech/Denoising-UNet/tree/main

Try it

Open the Space in your browser, or copy it to your own Hugging Face account from its page. Nothing in this kit runs by itself.